International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
furtherly refined by applying the Multi Station Adjustment
(MSA) procedure, a variant of the ICP (Iterative Closets
Point) algorithm implemented in Riegl's RiscanPro
companion software. This algorithm performs an initial
fitting of a set of planes in the point clouds and then tries to
align them by finding the best corresponding planes.
Because of the very large number of measurements acquired
with the Riegl VZ-400, merging and managing together both
full laser datasets revealed to be unfeasible. Therefore the
data analysis and comparison between the two laser scanners
were restricted to a limited area of the landslide, shown on
the right side of figure 2. To this aim a specific procedure
was adopted to merge together the corresponding scans. In a
first step the global registration was applied separately to
both datasets, then only the scans covering the selected area
were imported in a new RiscanPro project and furtherly
registered with the MSA plugin, in order to reduce as much
as possible residual misalignments. At the end of this
processing step we obtained a residual registration error of
about 3 cm (1 6) between the Z620 and VZ-400 selected
scans. This value is higher than the accuracy claimed by
Riegl for both laser scanners (10 mm for the Z620 and 5 mm
for the VZ-400). However it should be noted that the scans
were aligned prior to filter the vegetation, whose presence
may have therefore affected the fitting of planes and the
serach for correct correspondences between the scans during
the MSA procedure.
4. WAVEFORM PROCESSING IN TLS
Conventional terrestrial laser scanners based on the Time-Of-
Flight measurement principle characterize as analog discrete
return systems (Ullrich and Pfennigbauer, 2011). For each
emitted pulse, target detection and time-of-arrival (TOA)
estimation of the returned pulse are performed in real time
through analog devices. Regardless the various estimation
methods adopted, the resulting value of the TOA is affected
by trigger walk, i.e. by the amplitude of the target signal
detected by the receiver frontend. In presence of multiple
targets along the laser beam axis, analog estimators can yield
significant range errors for the second and further targets or
completely fail to detect them, depending on the temporal
separation between consecutive target echoes with respect to
the emitted pulse width. In contrast to a discrete return
systems, in an echo-digitizing system returned signals are
sampled at high rate and converted in a digital form prior to
perform the target detection. All subsequent processing steps
are executed in the digital domain on-line or in post-
processing. The latter approach is typically adopted in
airborne LiDAR systems where sample data are stored in
specific high capacity data recorders (Ullrich and Reichert,
2005). Applying the full-waveform analysis (FWA) to these
data enables to acquire additional information with respect to
conventional discrete return laser systems.
Beside range measurements, resulting from echo detection
and estimation of related TOA, backscattering properties of
the targets can be retrieved as well, such as the amplitude of
echo signal, which provides an estimate of target's laser
cross-section, and the pulse width, that represents a measure
of the backscatter profile of the target along the laser beam.
As mentioned in (Ullrich and Pfennigbauer, 2011), the
different approaches proposed so far to extract the target
backscattering properties from digitized returned signals can
be grouped into two main classes: deconvolution based
methods (Roncat et al., 2011) and procedures based on the
513
modeling of digitized waveform with basic functions
(Wagner et al., 2006; Roncat et al. 2008). An example of
the latter approach used for FWA is represented by the
Gaussian decomposition. This method relies on the
assumption that the system response can be modeled with a
Gaussian function and that all the contributions of the
backscattering targets are also Gaussian. Echo detection is
therefore performed by finding Gaussian pulses in the
returned waveform. Such approach has been implemented in
RiANALYZE, the Riegl software dedicated to the FWA of
echo-digittizing systems.
In recent years the Riegl company has developed a new line
of terrestrial laser scanners (VZ-series) providing a different
approach to FWA. In contrast to Airborne Laser Scanning
(ALS) systems, where digitized echo signals are stored
during the flight for subsequent post-processing, the lack of
computational power for real-time processing and the need to
immediately analyze the received signals, has led Riegl to
implement an online waveform processing for the VZ-line
products. Basically, upon echo pulse reception a highly
accurate estimate of its amplitude and TOA is performed in
real-time. Through hardware-oriented implementation of the
processing algorithm, a VZ-series laser scanner is able to
perform about 1.5 million range and amplitude measurements
per second. As denoted in table 2, given a laser pulse
repetition rate of 100 kHz (42000 measurements per second
in long range mode) and 300 kHz (125000 measurements per
second in high speed mode), the Riegl VZ-400 laser scanner
can record 10 or 5 targets per laser shot, respectively (Doneus
et al., 2009). Similarly to ALS-based echo-digitizing systems,
the Riegl's VZ-series instruments provide some additional
and very interesting features with respect to the conventional
analog discrete return-based terrestrial laser scanners, as
briefly described in the following subsections.
4.1 Multi-target capability
As previously mentioned, thanks to the adoption of online
waveform processing, the VZ-400 laser scanner can record
multiple echoes for each emitted laser pulse. However, the
capability to correctly discriminate two consecutive echoes is
determined by the laser's pulse width and the receiver
bandwidth: for the Riegl’s VZ-line laser scanners the multi-
target resolution (MTR) distance is about 0.8 m. Echo pulses
separated by shorter distances between scatterers within the
same laser shot cannot be distinguished as the corresponding
echo signals are overimposed. Consequently, the measured
range will be estimated somewhere in between the targets,
thus resulting in an erroneus point.
Since the processing of recorded waveform is performed in
real time and given the limited computational power
available on TLS systems, the Gaussian decomposition
method cannot be applied. This fact limits the multi-target
capability of VZ-series laser sensors and prevents to retrieve
the width of detected echoes as additional information.
However, the online waveform processing allows to reduce
the problem of too nearby targets by providing information
about the “pulse shape figure”. This parameter represents a
measure of the deviation of the actual target’s pulse shape
from the expected (and undistorted) pulse shape for each
individual echo. In this way, in cases where targets are closer
than the discrimination limit of 0.8 m, the pulse shape figure
allows to determine whether the return echo originates from a
single target or from at least two nearby targets.
